Act as a DevOps Engineer. Design a GitOps-based disaster recovery workflow for a stateless application running across two Kubernetes clusters (Region A and Region B) using ArgoCD. The goal is to achieve RTO (Recovery Time Objective) of under 15 minutes. Detail the repository structure required to manage the configuration for both regions. Explain how you would automate the failover process: if Region A becomes unreachable, how does the traffic routing (using a Service Mesh like Istio) automatically shift to Region B, and how do you ensure the Git state is reconciled correctly when Region A comes back online?